The current study examines the behaviors of four adverse impact measurements: the 4/5ths rule, two tests of significance (ZD and ZIR), and a newly developed AI measurement (Lnadj). Upon the suggestion of the Office of Federal Contract Compliance Program Manual about the sensitivity of the assessment of AI when the sample size is very large (Office of Federal Contract Compliance Programs, 2002), Lnadj is a new statistic that has been developed and proposed as an alternative practical significance test to the 4/5ths rule. The results indicated that, unlike the 4/5ths rule and other tests for adverse impact, Lnadj is an index of practical significance that is less sensitive to differences across selection conditions that are not supposed to affect tests of adverse impact. Furthermore, Lnadj decreases Type I error rates when there is a small d value and Type II error rates when there is moderate to large d value.
| Identifer | oai:union.ndltd.org:RICE/oai:scholarship.rice.edu:1911/64689 |
| Date | 06 September 2012 |
| Creators | Ercan, Seydahmet |
| Contributors | Oswald, Frederick L. |
| Source Sets | Rice University |
| Language | English |
| Detected Language | English |
| Type | thesis, text |
| Format | application/pdf |
Page generated in 0.0021 seconds